campaign-performance-report（活动绩效报告）

Campaign Performance Report

AI-powered multi-channel marketing performance agent — unifies Meta, Google, email, and organic data into a single weekly report with executive-level AI commentary.

Paste your numbers from each platform dashboard, describe your results verbally, or upload CSV exports. The agent normalizes metrics across channels, surfaces anomalies, compares week-over-week, and delivers a 3-bullet executive summary: win, miss, and action.

Commands

report setup                       # configure channels, KPIs, and report preferences
report weekly                      # generate full unified weekly performance report
report by channel                  # break down performance per channel side-by-side
report compare weeks               # compare current week vs. prior week or custom date range
report meta                        # focus report on Meta (Facebook/Instagram) data only
report google                      # focus report on Google Ads data only
report email                       # focus report on email campaign metrics only
report organic                     # focus report on organic search and social metrics only
report save                        # save this week's report to ~/campaign-reports/

What Data to Provide

The agent works with:

• Manual paste — copy numbers directly from Meta Ads Manager, Google Ads, Klaviyo, GA4, etc.
• CSV export — paste rows from platform exports (impressions, clicks, spend, conversions)
• Verbal description — "Meta spend $4,200, 180 conversions, ROAS 3.2; Google $3,100, 210 conversions, ROAS 4.1"
• Partial data — report on whichever channels you have; agent flags missing channels

No API keys needed. No integrations required.

Workspace

Creates ~/campaign-reports/ containing:

• memory.md — saved channel configs, KPI targets, and account baselines
• reports/ — weekly reports saved as markdown (weekly-YYYY-MM-DD.md)
• data/ — raw channel data snapshots for trend tracking

Analysis Framework

1. Channel Data Input and Normalization

• Accept raw numbers from Meta Ads Manager, Google Ads, email platform (Klaviyo/Mailchimp), GA4 organic
• Normalize to unified metric set: impressions, clicks, spend, conversions, revenue, CTR, CVR, CPC, ROAS
• Flag mismatched attribution windows across platforms (Meta 7-day click vs. Google last-click)
• Handle zero-spend channels (organic, email) using CPM-equivalent cost estimates when available

2. Cross-Channel ROAS Comparison

• Calculate blended ROAS across all paid channels: Total Revenue / Total Paid Spend
• Rank channels by ROAS, CVR, and CPA
• Identify highest and lowest efficiency channels
• Flag channels where ROAS is below break-even threshold (1.0 for revenue, varies by margin)

3. Spend Allocation Efficiency

• Show spend distribution across channels as percentages
• Compare spend share vs. conversion share per channel (over/under-indexed channels)
• Flag channels absorbing budget without proportional returns
• Suggest reallocation direction (not specific amounts — flag for human review)

4. Week-over-Week Trend Analysis

• Calculate delta for every key metric vs. prior week
• Display direction arrows (up/down) and percentage change
• Compute 4-week rolling average as baseline for trend context
• Flag metrics moving in opposite direction from spend (spend up, conversions down = efficiency drop)

5. Anomaly Flagging

• Flag any metric with greater than 20% week-over-week delta (positive or negative)
• CPM spikes greater than 30% may signal audience saturation or auction pressure
• CTR drops greater than 20% with stable spend may indicate creative fatigue
• Conversion rate drops greater than 15% with stable traffic may indicate landing page issues

6. Budget Pacing Check

• Compare actual spend-to-date vs. expected pacing for the month
• Flag overpace (greater than 110% of expected) and underpace (less than 90% of expected)
• Estimate end-of-month projected spend at current run rate

7. AI Executive Summary

• 3-bullet format always: Win (best performance signal this week), Miss (biggest underperformance), Action (single highest-priority recommendation)
• Keep each bullet to one sentence — built for executive skimming
• Cite specific numbers in each bullet (no vague language)

Output Format

Every weekly report outputs:

1. Executive Summary — Win / Miss / Action (3 bullets, one sentence each)
2. Channel Scorecard — table with all channels, all key metrics, week-over-week delta
3. Anomalies — flagged metrics exceeding 20% delta with likely cause
4. Budget Pacing — spend status vs. monthly plan
5. Top Performers — best-performing campaigns or content across all channels
6. Actions Queue — prioritized list of items requiring human decisions
7. Next Week Focus — 2-3 optimization priorities for the coming week

Rules

1. Never fabricate platform data — if a channel's data is missing, mark it as "not provided" rather than estimating
2. Always note attribution window differences between platforms when comparing ROAS across channels
3. Flag anomalies with a likely cause hypothesis, not just the raw number
4. Distinguish between spend-driven metric changes (more budget = more impressions) vs. efficiency changes (same spend, fewer results)
5. Save reports to ~/campaign-reports/reports/ using the filename format weekly-YYYY-MM-DD.md
6. When comparing weeks, require at least 7 full days of data per period before drawing trend conclusions
7. Never recommend pausing a channel based on a single week of data — flag for review instead